Spero Consulting

Recently, we’ve seen many op-​ed essays call­ing for more trans­parency in finan­cial state­ments, par­tic­u­larly with respect to mortgage-​related secu­ri­ties. Many of these essays have been writ­ten by famous and esteemed indi­vid­u­als or their staffs.

In our own idio­syn­cratic, round-​about way, we’ll explain the empty silli­ness of such argu­ments, and we begin by crit­i­ciz­ing the notion that “more is always better.”

Too Much Infor­ma­tion: Unfor­tu­nately, we’ve not read a sin­gle essay that con­tained an intel­li­gent, con­crete argu­ment for why more trans­parency is bet­ter than less – as if trans­parency, in and of itself, is a good (or is inher­ently good).

More pre­cisely, in all of these arti­cles, the value of trans­parency is assumed, and the assump­tion seems to be implicit and sub­con­scious (uncon­scious?) rather than some­thing arrived at via seri­ous delib­er­a­tion. (Hint: we can’t recall any of these essays that bother to define trans­parency. Pre­sum­ably, it is like pornog­ra­phy: you know it when you see it.)

In that half-​assed way, these recent prompts for more trans­parency have much in com­mon with the slightly older admo­ni­tions to elim­i­nate mark-​to-​market account­ing.¹

In their the­o­ries, many econ­o­mists – includ­ing, yours truly – have shown that more trans­parency, which often means more pre­cise infor­ma­tion, is not always bet­ter than less; in fact, it can make things strictly worse. Such seem­ingly patho­log­i­cal results are actu­ally rather com­mon in a vari­ety of social set­tings, includ­ing some markets, and arise for a num­ber of reasons, including risk-​sharing and incen­tives, where more infor­ma­tion can affect an agent’s behav­ior and actions or efforts thereby reduc­ing social wel­fare and/​or exac­er­bat­ing incen­tive problems.

For exam­ple (and this is a gross gen­er­al­iza­tion of the results with­out spec­i­fy­ing any of the nec­es­sary assumptions) in Kan­odia, Singh and Spero (JAR, 2005), we show that it is bet­ter to keep two unknown vari­ables as unknowns rather than know only one with per­fect pre­ci­sion. Think of it in the fol­low­ing way: sup­pose there are two ran­dom vari­ables – one that is some­what in the person’s con­trol and the other, which is not.

If the one under his influ­ence is known per­fectly, he’ll overem­pha­size it. If the other one is known per­fectly, then he’ll right­fully con­clude that the noisy sig­nal of his effort will be over­looked in favor of the other vari­able so he’ll do lit­tle. The for­mer cre­ates over-​exertion and the lat­ter cre­ates under-​exertion and both are socially dam­ag­ing; thus, one can find a happy medium in less extreme cases where nei­ther vari­able is known with total pre­ci­sion. (It should remind one of Goldilocks.)

Now, let’s be very clear that one need not be an econ­o­mist to know that more infor­ma­tion or trans­parency is not always bet­ter. For exam­ple, how does the reader answer ques­tions from a spouse, rel­a­tive, or friend when asked some­thing like, “Do you like my new hair­cut?” or “Does this dress make me look fat?”

In addi­tion, there are other cases where another party reveals per­sonal details with too much pre­ci­sion. In fact, we as a soci­ety have the col­lo­qui­al­ism, “Too much infor­ma­tion!” for just such cases where you’ll never again look at the revealer in the same man­ner and sub­se­quently rue­fully won­der, “why did they have to tell me that?”

Details Are Not Infor­ma­tion: this is a par­tic­u­larly apt time to repeat our admo­ni­tion that details are not infor­ma­tion. Back in April, we posted a long essay on the dif­fer­ence between details and infor­ma­tion or use­ful facts. (Use­ful facts are ones that might cause a deci­sion to change as the fact is real­ized.) Our point in that essay was to dis­tin­guish between keep­ing track of a lot of nec­es­sary data – as in data pro­cess­ing – and the quite dif­fer­ent task of pro­vid­ing use­ful infor­ma­tion to decision-makers. If one leaves sys­tems design to sys­tems peo­ple, one will likely get the for­mer and not much of the lat­ter. More­over, if the decision-​maker can’t design the sys­tem – not the pro­gram­ming – then his or her com­pe­tence at decision-​making should be jus­ti­fi­ably questioned.

The same dis­tinc­tion between details and infor­ma­tion holds true with finan­cial assets, too. More trans­parency can mean an inun­da­tion of book-​keeping and account details, which may pro­vide no infor­ma­tion or which may require expert judg­ment to (sift through to) become infor­ma­tion. In either case, the recip­i­ent of the data dump may not “see the for­est for the trees.“² So, one may have all the facts, but no abil­ity to orga­nize them – much like a child writ­ing a term paper.

And, that, of course, illus­trates the silli­ness of call­ing for more trans­parency for mortgage-​related secu­ri­ties. The big­ger prob­lem is that with every datum about every mort­gage in a pool, there is still no easy way to value them.

The issue isn’t the details, it is how to com­bine cur­rent and past details to deter­mine value and risk in the future, and it is very likely a per­fect method is unknow­able. So…

Value Matters, BUT There’s No Trans­par­ent Way to Find It: let’s illus­trate the notion in to a fairly high level of detail (for a blog post). We’ll ignore the “water­fall” aspect of real mortgage-​backed secu­ri­ties and CDOs where dif­fer­ent classes of secu­rity hold­ers have dif­fer­ent pri­or­ity claims on the cash flows because those claims are not the con­found­ing fac­tors – the intere­la­tion­ships of the mort­gages are.

So, imag­ine a pool of T thou­sand mort­gages going down the first col­umn of a spread­sheet. Fur­ther, sup­pose that the next 360 columns rep­re­sent months, m, so, the row t and col­umn m inter­sec­tion is the amount of cash received from bor­rower t in month m. Now that cell will actu­ally be a func­tion of any num­ber of fac­tors, includ­ing inter­est rates which affect whether the mort­gage is repaid early; the person’s wealth and income which deter­mine whether the bor­rower declares bank­ruptcy, the rela­tion­ship between the value of the col­lat­eral and the loan bal­ance, etc. We could go on and on, but the point is that each cell could take any num­ber of val­ues depend­ing upon many dif­fer­ent factors.

One page of the spread­sheet would then rep­re­sent one entire sce­nario of how cash is received from all T thou­sand mort­gages over the next thirty years.

At issue for val­u­a­tion (and risk mod­el­ing) is how to com­bine out­comes across all mort­gages. The cells are clearly related within a row, i.e., a borrower’s sta­tus in one month will affect cash flows in later months.

But, cash flows are also related within columns – phe­nom­ena, like a hur­ri­cane, may con­tem­po­ra­ne­ously affect more than one bor­rower – and across columns, too. For exam­ple, someone’s default in month m may make another’s default in month m + n more likely. So, the big­ger issue is: how does one relate bor­row­ers across time and space to arrive at a dis­tri­b­u­tion of cash flows. (Note: we mean “space” lit­er­ally because com­mu­nity and regional effects mat­ter – the inter-​row action, sometimes.)

One could gen­er­ate any num­ber of sce­nar­ios or pages, but, of course, the issue for val­u­a­tion (and risk) are which com­bi­na­tions in the numer­ous T x 360 grid are more (or less) likely (and how wide is the range of pos­si­ble outcomes)?

In other words, the prob­lem lays with deter­min­ing the joint dis­tri­b­u­tions across bor­row­ers and time. As we see it, there is no cor­rect method, but there is an infin­ity of incor­rect meth­ods, espe­cially ones that rely only on his­tor­i­cal rela­tion­ships, par­tic­u­larly very short histories.

Those incor­rect meth­ods include many that were imple­mented in recent years. As we see it, many of those meth­ods were imple­mented because they were solv­able, not because they were accu­rate. Unfor­tu­nately, those weak­nesses (inac­cu­ra­cies) were obscured by the rel­a­tive calm­ness of the mar­kets, includ­ing the near-​Ponzi-​like schemes of dif­fer­ent banks buy­ing the secu­ri­ties to re-​securitize them yet another time.

So, we ask those writ­ers urg­ing more trans­parency: exactly how would it help us find a price in the above exam­ple? Our illus­tra­tion high­lights the rea­son why there is a lack of buy­ers. There are data aplenty. What is lack­ing is a quan­tifi­able notion of the future.

That, dear reader, is why we devel­oped and wrote about an alter­na­tive solu­tion to TARP. One that involved the use of invest­ment tax cred­its or cash-​basis account­ing (to per­mit the imme­di­ate expense of the pur­chase price) to sub­si­dize and cush­ion the risk of pur­chas­ing these con­glom­er­a­tions of cash flows. It would pro­vide pri­vate buy­ers with an imme­di­ate ben­e­fit of 30% — 40% of the pur­chase price, which seems large enough to per­mit room for error.

As always, we encour­age vis­i­tors to read our essay, Uncer­tainty Man­age­ment, which dis­cusses the notions of mea­sur­a­bil­ity (quan­tifi­a­bil­ity) and immea­sur­a­bil­ity by dis­tin­guish­ing between the broader idea of uncer­tainty and the nar­rower idea of risk. In that regard, the num­ber and cost of mis-​specification errors related to our ongo­ing cri­sis may be the great­est in any period in history.

We’ll prob­a­bly edit this again in the near future.

Foot­notes: